A bias analysis on a breast cancer mammography dataset for deep learning applications

Blanca Zufiria Gerbolés, Karen López-Linares, María Jesús García González, Kristin May Rebescher, Marcos Rubio, Esther Albertin, David Chaparro, Maria Blanca Cimadevila, Javier García, Maria Jesús Ledesma-Carbayo, Iván Macía Oliver

Keywords: bias, deep learning, mammography, breast cancer.

Abstract: The development of democratized, generalizable deep learning applications for health care systems is challenging as potential biases could easily emerge. This paper provides an overview of the potential biases that appear in image analysis datasets that affect the development and performance of computer-aided algorithms. Furthermore, we summarize some techniques to alleviate these biases. Particularly, we focus on possible biases on a mammography dataset and we present a classification task to analyze the influence of biases in the performance of the algorithm.

Registration: I acknowledge that publication of this at MIDL and in the proceedings requires at least one of the authors to register and present the work during the conference.

Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.

Paper Type: methodological development

Primary Subject Area: Uncertainty Estimation

Secondary Subject Area: Application: Radiology

Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.